A method and system for evaluating and predicting the life of a generator slip ring carbon brush

Abstract

This invention discloses a method and system for assessing the condition and predicting the lifespan of a generator slip ring carbon brush, belonging to the field of generator operation, maintenance, and condition monitoring technology. The method includes: simultaneously acquiring temperature distribution data, friction acoustic emission signal data, and macroscopic visual image data of the carbon brush and slip ring, along with auxiliary detection data; analyzing carbon brush edge features to obtain carbon brush edge cracking rating results; extracting time-frequency domain feature parameters to construct a friction anomaly index characterizing the microscopic friction state; inverting the contact pressure drop at the carbon brush and slip ring interface; fusing the above multi-dimensional data to construct a multi-dimensional evaluation index system for the carbon brush health state; establishing a three-dimensional coupled attenuation nonlinear state-space model; and using an unscented Kalman filter algorithm, combined with the multi-dimensional evaluation index system, to recursively estimate the carbon brush health state and extrapolate the remaining lifespan and its confidence interval. This invention improves the accuracy of carbon brush condition assessment and the reliability of remaining lifespan prediction.

Description

Technical Field

[0001] This invention belongs to the field of generator operation, maintenance and condition monitoring technology, specifically relating to a method and system for assessing the condition and life prediction of generator slip ring carbon brushes. Background Technology

[0002] The generator slip ring-carbon brush system is the core actuator of the excitation system, responsible for introducing the stationary excitation power into the generator rotor windings through rotating slip rings to establish a rotating magnetic field. As a sliding electrical contact element, the carbon brush operates under extremely harsh conditions, subjected to high-speed rotation (typically linear velocities of 40-60 m / s), large currents (thousands of amperes), and spring pressure. The health of the carbon brush directly determines the quality of excitation current transmission and system stability. Statistics show that among various generator faults, shutdowns caused by carbon brush failures in the excitation system account for a significant proportion, often characterized by suddenness and severe damage.

[0003] Currently, the monitoring and maintenance of carbon brushes mainly involves manual inspections and periodic maintenance. Maintenance personnel, based on experience, periodically (e.g., daily or per shift) enter the slip ring chamber to measure carbon brush temperature using an infrared thermometer, visually observe the spark level, and feel the spring pressure by hand pressing the brush grip, or measure the remaining length of the carbon brush using measuring tools. This method cannot capture sudden deterioration between inspections; the determination of spark level varies from person to person, and consistency in temperature measurement points is difficult to guarantee; early signs of failure such as internal oxidation of the carbon brush, micro-cracks at the edges, and abnormal micro-friction are difficult to detect with the human eye and experience. For example, online monitoring of a single parameter is being explored. Some power plants are beginning to install online monitoring devices in the slip ring chamber, using thermocouples or integrated infrared probes embedded in the brush grip or carbon brush tail to achieve real-time acquisition of carbon brush temperature and over-limit alarms. Alternatively, industrial cameras can be used to photograph the slip ring area, and image processing algorithms can be used to identify the frequency and intensity of electrical sparks. In these types of solutions, the single-point arrangement of temperature monitoring cannot obtain the temperature distribution gradient; spark detection can only trigger an alarm when the fault has developed to a certain extent (obvious sparks appear), which is a post-event monitoring and cannot achieve early warning. Another example is the use of multi-parameter integrated monitoring and intelligent diagnosis. Some studies have begun to try to integrate data from multiple sensors, such as simultaneously collecting multiple signals such as current, voltage, temperature, and vibration, and classifying faults through neural networks.

[0004] However, currently, most assessment and prediction methods for generator slip ring carbon brushes neglect the microscopic frictional information of the carbon brush-slip ring friction pair. Frictional noises (such as chatter and stick-slip) often precede changes in electrical parameters and thermal effects, serving as precursors to failure, but existing technologies fail to effectively capture and utilize this information. Contact voltage drop is the most direct electrical parameter reflecting the state of the carbon brush-slip ring contact interface, but it is difficult to measure directly. Existing technologies mostly estimate it indirectly by measuring the brush braid voltage drop, lacking key parameter inversion, resulting in significant errors and failing to achieve accurate acquisition based on temperature field inversion. Therefore, overall, current assessment and prediction methods for generator slip ring carbon brushes suffer from incomplete multi-dimensional fusion of sensor information, lack of key parameter inversion, and difficulty in deeply revealing the carbon brush degradation mechanism. The accuracy of carbon brush condition assessment and the reliability of remaining life prediction need further optimization. Summary of the Invention

[0005] This invention provides a method and system for assessing the condition and predicting the lifespan of generator slip ring carbon brushes. The purpose is to address the current issues in the assessment and prediction of generator slip ring carbon brushes, such as incomplete fusion of multi-dimensional sensor information, lack of key parameter inversion, difficulty in deeply revealing the carbon brush degradation mechanism, and the need for further optimization of the accuracy of carbon brush condition assessment and the reliability of remaining life prediction.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a method for assessing the condition and predicting the lifespan of a generator slip ring carbon brush, comprising the following steps: S1. Simultaneously collect temperature distribution data, friction acoustic emission signal data, macro vision image data, and auxiliary detection data such as excitation current, remaining carbon brush length, and ambient temperature and humidity during the operation of carbon brush and slip ring. S2. Based on macro visual image data processing and analysis, carbon brush edge features are analyzed to obtain carbon brush edge cracking rating results; S3. Extract time-frequency domain feature parameters based on triboacoustic emission signal data to construct a friction anomaly index characterizing the micro-friction state; S4. Based on temperature distribution data and combined with the electrical and geometric parameters of the carbon brush, the contact pressure drop at the contact interface between the carbon brush and the slip ring is obtained by inversion. S5. By integrating carbon brush edge cracking rating, friction anomaly index, contact pressure drop, carbon brush remaining length, and carbon brush axial temperature gradient and oxidation degree index calculated based on temperature distribution data, a multi-dimensional evaluation index system for carbon brush health status is constructed. S6. Based on a multi-dimensional evaluation index system, establish a three-dimensional coupled decay nonlinear state-space model containing carbon brush wear, oxidation degree, and contact pressure drop; S7. Based on the three-dimensional coupled attenuation nonlinear state-space model, the unscented Kalman filter algorithm is used, combined with a multi-dimensional evaluation index system to recursively estimate the carbon brush health status, and extrapolate to obtain the remaining lifespan of the carbon brush and its confidence interval.

[0007] In some implementations, in S2, the processing and analysis of macro visual image data includes image preprocessing, semantic segmentation of the carbon brush region, and quantization of edge morphological features. The quantized features are then input into a classification model to obtain the carbon brush edge cracking rating result.

[0008] In some implementations, in S3, the processing of the frictional acoustic emission signal data first involves bandpass filtering, then frame division using a sliding time window. The extracted time-frequency domain feature parameters include root mean square value, ring count, peak factor, frequency centroid, and kurtosis. The frictional anomaly index is obtained by dimensionality reduction through principal component analysis to obtain the first principal component or by calculating Mahalanobis distance.

[0009] In some implementations, in S4, the inversion of the contact pressure drop is to solve the inverse problem of heat conduction, based on the carbon brush axial heat conduction equation: ; in, This is the distance from the contact surface between the carbon brush and the slip ring. This refers to the axial temperature distribution of the carbon brush. The perimeter of the carbon brush cross-section is... This represents the cross-sectional area of ​​the carbon brush. The heat flux density of the heat source inside the carbon brush. The convective heat transfer coefficient is... The thermal conductivity of the carbon brush material. For ambient temperature, The density of the carbon brush material. This refers to the specific heat capacity of the carbon brush material. For time.

[0010] In some implementations, the multidimensional evaluation index system constructed in S5 is in the form of a multidimensional vector of carbon brush health status, as follows: ; in, A multidimensional vector representing the health status of carbon brushes. This represents the remaining length of the carbon brush. This is an index of the degree of carbon brush oxidation. This refers to the contact pressure drop at the interface between the carbon brush and the slip ring. The axial temperature gradient of the carbon brush over continuous time. This is the friction anomaly index. This refers to the carbon brush edge chipping level. For time.

[0011] In some implementations, in S6, the discrete-time state equation of the three-dimensional coupled decaying nonlinear state-space model is: ; in, This refers to the amount of carbon brush wear. The degree of carbon brush oxidation, This refers to the contact voltage drop between the carbon brush and the slip ring. For excitation current, The linear velocity of the slip ring. This refers to the carbon brush edge chipping level. This refers to the core temperature of the carbon brush. As the reference temperature, This represents the total length of the carbon brush. For spring pressure, The initial pressure of the spring. The sampling interval is... For discrete-time indexing, The parameters to be identified in the model This is process noise.

[0012] In some implementations, in S6, the observation equation of the three-dimensional coupled decaying nonlinear state-space model is: ; in, For discrete-time indexing, This represents the remaining length of the carbon brush in discrete time. This is an index of the degree of carbon brush oxidation over discrete time. These are the observed contact pressure drops between the carbon brush and the slip ring over discrete time periods. The axial temperature gradient of the carbon brush in discrete time. This represents the friction anomaly index in discrete time. The carbon brush edge chipping level is defined as a discrete-time condition. This represents the carbon brush wear over discrete time intervals. The degree of carbon brush oxidation at discrete time. The contact voltage drop between the carbon brush and the slip ring is given in discrete time. The excitation current is in discrete time. It is a nonlinear function. To observe noise.

[0013] In some implementations, when recursively estimating the carbon brush health status using the unscented Kalman filter algorithm in S7, Sigma points are first generated based on the current state estimate and covariance. The calculation formula is as follows: ; in, for The initial Sigma point at time t, for The first moment Sigma points, for Estimated carbon brush health status at any given time. Let be the state dimension. For scale parameters, for The covariance matrix at time t, The first square root of the matrix List.

[0014] In some implementations, in S7, the extrapolation of the remaining lifespan of the carbon brush includes: comparing the predicted state trajectory with a preset failure threshold, the failure threshold including wear failure, electrical failure and thermal aging failure, wear failure being the carbon brush wear reaching its maximum value, electrical failure being the contact voltage drop reaching its maximum value, and thermal aging failure being the oxidation degree reaching its maximum value; multiplying the difference between the discrete time when the state variable first crosses any failure threshold and the current discrete time by the sampling interval to obtain the remaining lifespan; and generating multiple predicted trajectories through Monte Carlo simulation to obtain the confidence interval of the remaining lifespan.

[0015] This invention also provides a system for assessing the condition and predicting the lifespan of generator slip ring carbon brushes, used to implement the aforementioned method for assessing the condition and predicting the lifespan of generator slip ring carbon brushes. The system includes a multi-source heterogeneous data acquisition module, an automatic rating module for carbon brush edge chipping, a friction noise feature extraction module, a contact pressure drop and temperature field inversion module, a multi-dimensional evaluation system construction module, a three-dimensional attenuation model modeling module, and a dynamic prediction module for remaining lifespan. The multi-source heterogeneous data acquisition module is used to: synchronously acquire temperature distribution data, friction acoustic emission signal data, macro vision image data, and auxiliary detection data such as excitation current, remaining carbon brush length, and ambient temperature and humidity during the operation of carbon brush and slip ring. The automatic carbon brush edge chipping rating module is used to: analyze carbon brush edge features based on macro visual image data processing and obtain carbon brush edge chipping rating results; The friction noise feature extraction module is used to: extract time-frequency domain feature parameters based on friction sound emission signal data, and construct a friction anomaly index characterizing the micro-friction state; The contact pressure drop and temperature field inversion module is used to: invert the contact pressure drop at the contact interface between the carbon brush and the slip ring based on temperature distribution data combined with the electrical and geometric parameters of the carbon brush; The multidimensional evaluation system construction module is used to: integrate carbon brush edge cracking rating, friction anomaly index, contact pressure drop, carbon brush remaining length, and carbon brush axial temperature gradient and oxidation degree index calculated based on temperature distribution data to construct a multidimensional evaluation index system for carbon brush health status. The three-dimensional attenuation modeling module is used to: establish a three-dimensional coupled attenuation nonlinear state-space model containing carbon brush wear, oxidation degree, and contact pressure drop based on a multi-dimensional evaluation index system; The remaining life dynamic prediction module is used to: recursively estimate the carbon brush health status based on a three-dimensional coupled decay nonlinear state-space model, using an unscented Kalman filter algorithm, combined with a multi-dimensional evaluation index system, and extrapolate to obtain the remaining life of the carbon brush and its confidence interval.

[0016] Compared with the prior art, the present invention provides a method and system for evaluating the condition and predicting the lifespan of a generator slip ring carbon brush, which has the following advantages: This invention provides a method for assessing the condition and predicting the lifespan of generator slip ring carbon brushes. It introduces an acoustic emission sensor into slip ring carbon brush condition monitoring, enabling the capture of microscopic frictional abnormalities (such as stick-slip, abrasive wear, and microcrack initiation) that are imperceptible by traditional methods, achieving ultra-early warning of faults. This invention constructs a comprehensive evaluation system based on multi-dimensional information fusion. Through the complementarity of multi-source heterogeneous data, it improves the accuracy and robustness of condition assessment, effectively avoiding false alarms and missed alarms. A three-dimensional coupled attenuation model is proposed, establishing a nonlinear state-space model describing the mutual coupling of wear, oxidation, and electrical contact. This model truly reflects the physical nature of carbon brush degradation, providing a solid mechanistic basis for remaining life prediction. By combining the nonlinear state-space model with the unscented Kalman filter algorithm, real-time observation data is fully utilized to perform online correction of the model state and predict the state trajectory forward until failure, thereby dynamically outputting the remaining lifespan of the carbon brush and its confidence interval. Compared to methods based on empirical formulas or simple thresholds, the prediction accuracy and reliability of this invention are greatly improved, providing a scientific basis for implementing predictive maintenance, optimizing spare parts inventory, and rationally arranging shutdown and maintenance plans. It can be applied in the existing network environment of power plants and has a certain degree of engineering adaptability. Attached Figure Description

[0017] The accompanying drawings are provided to further understand the invention and constitute a part of this invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0018] Figure 1 This is a flowchart illustrating the method for evaluating the condition and predicting the lifespan of a generator slip ring carbon brush according to the present invention. Figure 2 This is a schematic diagram of the installation layout of multi-source sensors in the slip ring chamber in an embodiment of the generator slip ring carbon brush condition assessment and life prediction method of the present invention. Figure 3 This is a flowchart of an automatic rating method for carbon brush edge chipping based on deep learning, as part of an embodiment of a generator slip ring carbon brush condition assessment and life prediction method of the present invention. Figure 4This is a schematic diagram illustrating the extraction of acoustic emission signal features and the construction of a friction anomaly index in an embodiment of a generator slip ring carbon brush condition assessment and life prediction method of the present invention; wherein, Figure 4 (a) is a schematic diagram of the original acoustic emission signal; Figure 4 (b) is a schematic diagram showing the trend of characteristic parameter changes; Figure 5 This is a schematic diagram of the contact pressure drop inversion principle using a finite element and neural network surrogate model, as part of an embodiment of the generator slip ring carbon brush condition assessment and life prediction method of the present invention. Figure 6 This is a flowchart of the remaining lifetime prediction algorithm based on unscented Kalman filtering in an embodiment of the generator slip ring carbon brush condition assessment and lifetime prediction method of the present invention. Figure 7 In an embodiment of the method for assessing the condition and predicting the lifespan of a generator slip ring carbon brush according to the present invention, a visual interface is displayed showing a carbon brush health radar chart and a Gantt chart of remaining life. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0022] It should be noted that the apparatus and methods disclosed in the embodiments herein can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments herein. In this regard, each block in a flowchart or block diagram may represent a module, program, or part of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system to perform the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0023] In addition, the functional modules in the various embodiments of this article can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0024] How can we provide a method that can integrate multi-dimensional sensing information, deeply reveal the degradation mechanism of carbon brushes, and have high-precision remaining life prediction capabilities, so as to achieve predictive intelligent maintenance of generator slip ring carbon brushes?

[0025] Based on this, the present invention provides a method for evaluating the condition and predicting the lifespan of a generator slip ring carbon brush, comprising the following steps: S1. Simultaneously collect temperature distribution data, friction acoustic emission signal data, macro vision image data, and auxiliary detection data such as excitation current, remaining carbon brush length, and ambient temperature and humidity during the operation of carbon brush and slip ring. S2. Based on macro visual image data processing and analysis, carbon brush edge features are analyzed to obtain carbon brush edge cracking rating results; S3. Extract time-frequency domain feature parameters based on triboacoustic emission signal data to construct a friction anomaly index characterizing the micro-friction state; S4. Based on temperature distribution data and combined with the electrical and geometric parameters of the carbon brush, the contact pressure drop at the contact interface between the carbon brush and the slip ring is obtained by inversion. S5. By integrating carbon brush edge cracking rating, friction anomaly index, contact pressure drop, carbon brush remaining length, and carbon brush axial temperature gradient and oxidation degree index calculated based on temperature distribution data, a multi-dimensional evaluation index system for carbon brush health status is constructed. S6. Based on a multi-dimensional evaluation index system, establish a three-dimensional coupled decay nonlinear state-space model containing carbon brush wear, oxidation degree, and contact pressure drop; S7. Based on the three-dimensional coupled attenuation nonlinear state-space model, the unscented Kalman filter algorithm is used, combined with a multi-dimensional evaluation index system to recursively estimate the carbon brush health status, and extrapolate to obtain the remaining lifespan of the carbon brush and its confidence interval.

[0026] This invention simultaneously acquires temperature, friction acoustic emission signals, and visual images of carbon brushes using an infrared thermal imager, acoustic emission sensor, and macro vision sensor. It utilizes a deep learning semantic segmentation network for image processing to achieve automatic rating of carbon brush edge chipping. Time-frequency domain features of the acoustic emission signals are extracted to construct a friction anomaly index. Contact pressure drop is inverted based on the temperature field. Furthermore, a comprehensive health status evaluation system is constructed by integrating multi-dimensional indicators such as wear, oxidation, and contact pressure drop. A nonlinear state-space model describing the coupled evolution of wear, oxidation, and contact pressure drop is established, and an unscented Kalman filter algorithm is used to dynamically predict remaining life. This invention applies acoustic emission technology to carbon brush micro-friction monitoring and constructs a three-dimensional coupled attenuation model, improving the accuracy of carbon brush condition assessment and the reliability of remaining life prediction, providing key technical support for predictive maintenance of generator excitation systems.

[0027] like Figure 1 As shown, specifically, the method of the present invention is carried out according to the following steps: S1: Synchronous Acquisition of Multi-Source Heterogeneous Data Install and fix the following sensors at predetermined locations on the inner wall of the generator slip ring chamber: 1) An infrared focal plane detector is preferred, with a resolution of no less than 320×240 pixels, a temperature measurement range of -20℃ to 500℃, and a frame rate of ≥25Hz. Its installation angle should ensure that each carbon brush and slip ring surface is within the field of view without overlapping or obstruction.

[0028] 2) A wideband (100kHz-1MHz) resonant acoustic emission sensor with a built-in preamplifier (40dB gain) is selected. The sensor is tightly attached to the outside of the brush holder or the back of the carbon brush using a magnetic mount or a special clamp, and a coupling agent is used to ensure good acoustic coupling. The sensor is used to collect the elastic stress wave signal generated when the carbon brush slides relative to the slip ring.

[0029] 3) It adopts an industrial-grade CMOS camera with a resolution of no less than 5 megapixels, equipped with a long working distance macro lens and a ring LED supplementary light source. The lens is aimed at the edge and contact area of ​​the carbon brush through a quartz glass observation window on the slip ring chamber wall to obtain clear macro images.

[0030] 4) Includes Hall current sensor (for measuring excitation current), displacement sensor or laser rangefinder (for measuring remaining carbon brush length), and temperature and humidity sensor (for measuring ambient temperature and humidity).

[0031] All sensors receive a unified timing signal (such as GPS or PTP) through a synchronization trigger, enabling microsecond-level synchronous acquisition and ensuring strict time alignment of multi-source data.

[0032] S2: Automatic rating of carbon brush edge chipping This involves transforming the implicit mechanical damage information of carbon brushes in macro vision images into quantifiable health indicators. Specifically, this includes the following: S21. The acquired raw RGB images are converted to grayscale and histogram equalized to enhance contrast, and median filtering is used to remove noise. Due to the possible vibration of the slip ring chamber, the image sequence needs to be electronically stabilized, and inter-frame displacement is eliminated through feature point matching and affine transformation.

[0033] S22. Construct and train a deep learning-based semantic segmentation network to extract rich image features; the decoder recovers details through upsampling and skip connections, and finally outputs a probability map of each pixel belonging to the brush, slip ring, or background. The training dataset consists of manually annotated historical images, covering brush images with different wear levels and lighting conditions.

[0034] S23. Based on the segmented carbon brush mask, extract the carbon brush contour. Use the Canny edge detection operator to obtain fine edge lines. Then calculate the following morphological parameters: Defect area ratio Register and align the current brush outline with the standard brush template (the outline of a brand new brush), and calculate the pixel area of ​​the outline deviation area (defective area). A defect Total projected area of ​​carbon brush The ratio: ; 2) Maximum crack length The edge image is skeletonized, and crack-like lines are identified through 8-neighbor connected component analysis. The pixel length of the longest line is calculated. Combined with the known pixel-to-millimeter conversion factor, it is converted into the actual physical length (unit: mm).

[0035] 3) Number of broken pieces : Statistically identify independent defect areas (i.e., areas larger than a preset threshold) The number of connected components.

[0036] S24. Fracture Level Assessment: Using the above-mentioned quantitative features as input, a pre-trained Support Vector Machine (SVM) multi-classifier model is employed to output the fracture level of the carbon brush edge. C L The levels are defined as follows: Level 0 (Normal): ,and ,and .

[0037] Level 1 (Slight): , or exist A single, minute crack.

[0038] Level 2 (Moderate): , or exist Cracks, or .

[0039] Level 3 (Severe): ,or ,or .

[0040] S3: Extraction and Analysis of Friction Noise Features The microscopic frictional state between the carbon brush and the slip ring is precisely characterized using acoustic emission signals. Specifically, this includes: S31. Signal preprocessing: processing the raw voltage signal acquired by the acoustic emission sensor. Bandpass filtering (e.g., through a 100kHz high-pass filter and a 1MHz low-pass filter) is used to remove low-frequency mechanical vibration interference and high-frequency electromagnetic noise.

[0041] S32. Feature parameter calculation: using a sliding time window (window length) sliding step size The signal is processed by framing. For each frame, characteristic parameters are calculated: root mean square (RMS) value, ring count (count) (statistical signal waveform exceeding a preset threshold). (number of times), peak factor C F. Frequency centroid F c. kurtosis K .

[0042] S33. Construction of Friction Anomaly Index: The above-mentioned multiple feature parameters are integrated into a comprehensive friction anomaly index. Principal component analysis (PCA) can be used to reduce the dimensionality of the eigenvectors, and the first principal component can be taken as... Alternatively, statistical thresholds for each feature parameter can be established based on historical normal data, and the current feature vector can be defined by calculating the distance of its deviation from the normal state (such as Mahalanobis distance). . The larger the value, the more abnormal the micro-friction state.

[0043] S4: Contact Pressure Drop and Temperature Field Inversion The contact pressure drop at the interface between the carbon brush and the slip ring can be indirectly estimated by measuring the surface temperature distribution. Since contact voltage drop directly reflects the magnitude of contact resistance, it is a key parameter for evaluating contact condition and the degree of heat generation.

[0044] In some alternative implementations, the present invention is based on the inversion of an analytical thermal circuit model.

[0045] Assuming the carbon brush is a one-dimensional heat-conducting rod and neglecting the lateral temperature difference, the heat conduction equation along the axial direction of the carbon brush (from the contact surface to the direction of the brush braid) can be established: ; in, This is the distance from the contact surface between the carbon brush and the slip ring. This refers to the axial temperature distribution of the carbon brush. The perimeter of the carbon brush cross-section is... This represents the cross-sectional area of ​​the carbon brush. The convective heat transfer coefficient is... The thermal conductivity of the carbon brush material. For ambient temperature, The density of the carbon brush material. This refers to the specific heat capacity of the carbon brush material. For time, The internal heat source (Joule heating) is located at the contact interface. This is the sum of frictional heat and contact resistance heat. Under steady-state conditions, neglecting the time term, and considering the boundary conditions (heat flux density at the contact surface)... ,in The coefficient of friction, To contact pressure, (where the linear velocity is used), the carbon brush surface temperature can be derived. With contact pressure drop The approximate analytical relationship can be obtained by measuring the temperature at different locations on the carbon brush (e.g., using an infrared thermal imager to obtain the axial temperature distribution curve) and fitting this relationship using the least squares method. .

[0046] S5: Construction of a Multidimensional Evaluation Index System By integrating the key indicators obtained above, a multi-dimensional vector of the current health status of the carbon brush is formed. : ; in, A multidimensional vector representing the health status of carbon brushes. For time, The remaining length of the carbon brush (obtained by laser ranging or image measurement), in mm; The carbon brush oxidation degree index is defined as follows: ; in This is the core temperature of the carbon brush (which can be the highest temperature measured by infrared thermography). This is the temperature threshold at which the carbon brush material begins to oxidize significantly (determined through material testing). This index reflects the cumulative effect of high-temperature thermal aging on the carbon brush. The contact pressure drop at the carbon brush-slip ring interface, obtained from the S4 inversion, is expressed in mV. The axial temperature gradient of the carbon brush over continuous time is defined as the maximum difference in the axial temperature of the carbon brush, reflecting the current skin effect and the degree of local overheating. The friction anomaly index is obtained from step S3; The chipping level of the carbon brush edge is obtained from S2 (with values ​​of 0, 1, 2, 3).

[0047] S6: Modeling of a 3D Coupling Attenuation Model A core state-space model is constructed to describe the evolution of carbon brush health over time. Three mutually coupled state variables that best characterize the nature of carbon brush degradation are selected: wear amount. (Weared length, initial value 0, maximum value) ), degree of oxidation Contact pressure drop .

[0048] .

[0049] State equations (nonlinear, discrete time), let For discrete-time indexing, the sampling interval is... .

[0050] ; in, The excitation current is (A). The linear velocity of the slip ring (m / s) is calculated from the rotational speed. This represents the fracture level (discrete value). The core temperature of the carbon brush (°C) is obtained from infrared thermography. This is the reference temperature (°C). Below this temperature, oxidation acceleration is not considered. This refers to the amount of carbon brush wear. The degree of carbon brush oxidation, This refers to the contact pressure drop between the carbon brush and the slip ring. The spring pressure (N) can be obtained from a pressure sensor or displacement-pressure conversion. The initial pressure of the spring is (N). This represents the total length of the carbon brush. : indicates positive, that is . : The model parameters to be identified, which are positive real numbers. Process noise, assumed to be zero-mean Gaussian white noise.

[0051] Wear evolution: First item This represents wear caused by frictional work (Archard wear model form), where... The contact resistance heat power is positively correlated with the frictional work; the second term This indicates the inhibitory effect of oxidation on wear (the oxide film has a certain lubricating and protective effect, but excessive oxidation will lead to film peeling; here it is simplified to an exponentially decaying effect); the third term This indicates the accelerating effect of edge chipping on wear.

[0052] Evolution of oxidation degree: Oxidation is a temperature-driven diffusion process. (First item) This indicates the oxidation rate after the temperature exceeds the threshold; the second term... This indicates additional thermal aging caused by Joule heating; the third item This indicates a decrease in the degree of oxidation caused by the peeling or wear of the oxide layer (i.e., the oxide layer is worn away).

[0053] Contact pressure drop evolution: Contact pressure drop is affected by mechanical pressure and oxide film. (First item) This reflects the decrease in pressure due to spring elongation as the carbon brush wears, thus increasing the contact pressure drop; the second item This indicates that increased oxide film thickness increases contact resistance; the third item This indicates the self-recovery or regulation effect of the contact voltage drop (such as the softening effect of current heating at the contact point).

[0054] Observation equation: combining state variables with observable multidimensional indices They are connected. Some indicators can be directly regarded as observations of the state, while others are functions of the state.

[0055] ; in, For discrete-time indexing, This represents the remaining length of the carbon brush in discrete time. This is an index of the degree of carbon brush oxidation over discrete time. These are the observed contact pressure drops between the carbon brush and the slip ring over discrete time periods. The axial temperature gradient of the carbon brush in discrete time. This represents the friction anomaly index in discrete time. The carbon brush edge chipping level is defined as a discrete-time condition. This represents the carbon brush wear over discrete time intervals. The degree of carbon brush oxidation at discrete time. The contact voltage drop between the carbon brush and the slip ring is given in discrete time. The excitation current in discrete time; It is a nonlinear function, which can be obtained through mechanism or data fitting. For example, It can be derived based on the heat conduction model. To observe noise.

[0056] Model parameters It can be obtained through offline identification (optimization using historical lifecycle data) or online adaptive updating (such as using extended Kalman filtering to estimate state and parameters simultaneously).

[0057] S7. Dynamic prediction of remaining lifetime based on unscented Kalman filtering Since both the state equation and the observation equation are nonlinear, this invention employs unscented Kalman filtering (UKF) for state estimation and prediction. UKF approximates the distribution of the random variable after the nonlinear transformation through deterministic sampling (unscented transformation), avoiding the problems of Extended Kalman Filtering (EKF), which requires calculating the Jacobian matrix and suffers from large truncation errors. Specifically: S71. Initialization: exist At time , given the initial state estimate Covariance Matrix , where the state vector .

[0058] S72, For each moment Execute the following loop: a. Calculate the Sigma point: Based on the current state estimate Covariance ,generate Sigma points ( Here, the dimension of the state is... ): ; in, for The initial Sigma point at time t, for The first moment Sigma points, for Estimated carbon brush health status at any given time. Let be the state dimension. For scale parameters, for The covariance matrix at time t. It is a scale parameter. Determine the dispersion of Sigma points (usually taken as...) ), For secondary scaling parameters (usually taken as...) or ), Used to incorporate prior distribution information (optimal for Gaussian distribution). ). The first square root of the matrix List.

[0059] b. Time Update (Prediction): Substitute each Sigma point into the state equation (i.e., the discrete state equation in step S6) and perform a nonlinear transformation: ; in, For known input (e.g.) wait).

[0060] Calculate the predicted state mean and covariance: ; The weights are: ; c. Observation Update (Correction): The predicted Sigma point is passed through the observation equation. Transform into predicted observations: ; Calculate the predicted observation mean, covariance, and cross-covariance: ; Calculate Kalman gain : ; When new actual observations are obtained Then, update the state estimate and covariance: ; S73. Remaining life prediction: Obtain the current moment State Optimal Estimation Then, multi-step predictions of the future are performed. Using a forward simulation approach, starting from the current state, the future is iteratively calculated using the state equation (without noise terms). Step state trajectory In each prediction step, future inputs (such as current and speed) can be extrapolated from current values ​​or given based on the predicted load curve. Simultaneously, a failure threshold is defined: Wear and tear failure: (i.e., remaining length) ).

[0061] Electrical failure: .

[0062] Thermal aging failure: .

[0063] The predicted state trajectory is compared with the thresholds to find the moment when the state variable first crosses any threshold. Then the remaining lifespan Due to process noise and parameter uncertainties, multiple predicted trajectories can be generated through Monte Carlo simulation, thereby obtaining the probability distribution of RUL (such as mean, 5th percentile, 95th percentile), providing confidence intervals for maintenance decisions.

[0064] Based on the above prediction results, subsequent visualization and alerts can be implemented as needed.

[0065] The current health status (multi-dimensional indicators, overall health), remaining life prediction results (and confidence intervals), and historical trend curves of each carbon brush are dynamically displayed on the monitoring interface in the form of dashboards, radar charts, Gantt charts, etc. An alarm of the corresponding level is triggered when any of the following conditions are met: Warning: Heaven, or Exceeding the upper limit of the normal range, or the fracture level reaches level 2.

[0066] Call the police: Heaven, or Exceeding the threshold, or the fracture level reaches level 3.

[0067] Emergency shutdown: In case of severe temperature surge, arcing, or other extreme situations (can be linked with other protection systems).

[0068] This invention also provides a system for assessing the condition and predicting the lifespan of generator slip ring carbon brushes, including a multi-source heterogeneous data acquisition module, an automatic rating module for carbon brush edge chipping, a friction noise feature extraction module, a contact pressure drop and temperature field inversion module, a multi-dimensional evaluation system construction module, a three-dimensional attenuation model modeling module, and a remaining lifespan dynamic prediction module, wherein: The multi-source heterogeneous data acquisition module is used to: synchronously acquire temperature distribution data, friction acoustic emission signal data, macro vision image data, and auxiliary detection data such as excitation current, remaining carbon brush length, and ambient temperature and humidity during the operation of carbon brush and slip ring. The automatic carbon brush edge chipping rating module is used to: analyze carbon brush edge features based on macro visual image data processing and obtain carbon brush edge chipping rating results; The friction noise feature extraction module is used to: extract time-frequency domain feature parameters based on friction sound emission signal data, and construct a friction anomaly index characterizing the micro-friction state; The contact pressure drop and temperature field inversion module is used to: invert the contact pressure drop at the contact interface between the carbon brush and the slip ring based on temperature distribution data combined with the electrical and geometric parameters of the carbon brush; The multidimensional evaluation system construction module is used to: integrate carbon brush edge cracking rating, friction anomaly index, contact pressure drop, carbon brush remaining length, and carbon brush axial temperature gradient and oxidation degree index calculated based on temperature distribution data to construct a multidimensional evaluation index system for carbon brush health status. The three-dimensional attenuation modeling module is used to: establish a three-dimensional coupled attenuation nonlinear state-space model containing carbon brush wear, oxidation degree, and contact pressure drop based on a multi-dimensional evaluation index system; The remaining life dynamic prediction module is used to: recursively estimate the carbon brush health status based on a three-dimensional coupled decay nonlinear state-space model, using an unscented Kalman filter algorithm, combined with a multi-dimensional evaluation index system, and extrapolate to obtain the remaining life of the carbon brush and its confidence interval.

[0069] The present invention will be further described in detail below through specific embodiments.

[0070] This embodiment uses the slip ring carbon brush system of a 600MW steam turbine generator in a power plant as an example to implement the method of the present invention. The unit has a total of 48 carbon brushes in the slip ring chamber, 24 for the positive pole and 24 for the negative pole, with a rated excitation current of 2000A and a slip ring linear velocity of 40m / s.

[0071] like Figure 2 As shown, three infrared thermal imagers (320×256 pixels resolution, temperature range -20℃ to 500℃) are installed at equal intervals on the circumference of the slip ring chamber inner wall to ensure that the sides of each carbon brush and the surface of the slip ring are covered. A broadband acoustic emission sensor (frequency response 100kHz-1MHz, built-in 40dB preamplifier) ​​is installed on the side of each brush holder using a special fixture, for a total of 48 sensors. Two observation windows are opened on each side of the slip ring chamber, and four high-resolution industrial cameras (5 megapixels, equipped with a 50mm macro lens and a ring LED light source) are installed, aligned with the carbon brush array, to acquire macro images of the carbon brush edges. In addition, a Hall current sensor (measuring excitation current, range 0-3000A, accuracy 0.5%) is installed at the slip ring shaft end, a laser displacement sensor (range 0-50mm, accuracy 0.01mm) is installed above each brush holder to measure the remaining length of the carbon brush, and a temperature and humidity sensor is installed both inside and outside the slip ring chamber.

[0072] All sensors are synchronized via synchronous triggers, achieving microsecond-level synchronous data acquisition. The data acquisition system acquires infrared thermal image data (each pixel corresponds to a temperature value) at a rate of 1000 frames per second, acoustic emission signals (24-bit resolution) at a rate of 500 kS / s, and macro images (RAW format) at a rate of 10 frames per second. The data is transmitted in real time via fiber optic cable to a high-performance server (equipped with a GPU accelerator card) located among the electronic devices.

[0073] like Figure 3 As shown, the server calls the semantic segmentation model based on the improved U-Net every 5 minutes for the latest macro image of each carbon brush. This model was trained on 2000 manually annotated carbon brush images, achieving a segmentation accuracy (IoU) of 92% on the test set. Taking the image processing of carbon brush #23 as an example: 1) After image preprocessing, the carbon brush area is segmented and the edge contour is extracted.

[0074] 2) Compared with the standard template, the calculated defect area ratio is 3.2%, the maximum crack length is 1.5 mm, and the number of broken pieces is 2.

[0075] 3) Input the support vector machine classifier and output a collapse level of 2 (moderate collapse).

[0076] 4) The system records the result and marks the status of the carbon brush on the visualization interface.

[0077] like Figure 4 As shown, the acoustic emission signal is bandpass filtered (100kHz-1MHz) and then processed in frames with a window size of 10ms and a step size of 2ms. Figure 4 Image (a) shows a segment of the raw acoustic emission waveform (1 ms in duration) detected during a particular event, with a noticeable impulse component appearing at 0.5 ms. Calculate the following for this frame of signal: Root mean square (RMS) value = 0.12V (normal baseline is 0.05V); kurtosis K =8.5 (normally around 3.0); frequency centroid F C = 320 kHz (normally 150 kHz).

[0078] This indicates the presence of abnormal friction. For example... Figure 4 As shown in (b), the system integrates features such as RMS, kurtosis, frequency centroid, and ring count into a friction anomaly index through principal component analysis. The result of this calculation is... =4.2, exceeding the warning threshold of 3.0, the system issued a micro-friction anomaly warning. Maintenance personnel found that the pressure of the #23 carbon brush spring was too low (from the initial 25N to 18N). After adjustment... The situation gradually returned to normal, preventing potential fire accidents in the future.

[0079] like Figure 5 As shown, to achieve the inversion of contact pressure drop, a three-dimensional thermo-electric coupling model of the carbon brush-slip ring contact was first established in finite element software. The carbon brush has geometric dimensions of 32mm × 25mm × 100mm, and its material properties (thermal conductivity) are as follows: λ =15W / (mK), resistivity ρ =30μΩm, all linearly varying with temperature (determined experimentally). 5000 sets of input parameter combinations were generated through sampling: Excitation current I ∈[500, 2500]A, contact resistance R c∈[0.1,10]mΩ, convective heat transfer coefficient h ∈[10,50]W / (m²K).

[0080] For each set of parameters, the steady-state temperature field is solved, and the temperature values ​​of six feature points (axially distributed) on the carbon brush surface are extracted. A three-layer backpropagation neural network is trained using this data (input: I , R c, h Output: Temperature at 6 feature points; Hidden layer neurons: 20 and 10 respectively; Training error MSE < 0.1. This surrogate model can quickly and positively predict the temperature field.

[0081] When running online, with the current excitation current I =1850A and infrared thermal imager measured the temperature at 6 points on the surface of #23 carbon brush. Given a quantity, the contact resistance is obtained by solving an optimization problem. =1.2mΩ, and then the contact voltage drop is calculated. =2.22V. This value is higher than the normal value (1.5V), indicating that the contact condition has deteriorated.

[0082] The key indicators obtained above are integrated into a multi-dimensional vector of carbon brush health status. ,in: The remaining length of the carbon brush, measured by a laser displacement sensor, is currently 38 mm (the new carbon brush is 50 mm).

[0083] Oxidation degree index, calculated based on the cumulative core temperature exceeding 150°C, currently valued at 3.2 (dimensionless).

[0084] Contact voltage drop, obtained from inversion, is currently 2.22V.

[0085] Temperature gradient, calculated from infrared thermal imaging data, currently 20℃ (the difference between the maximum and minimum values).

[0086] Friction anomaly index, currently 4.2.

[0087] : Rupture level, currently 2.

[0088] These metrics are updated every minute and serve as the observation input for subsequent UKF filters.

[0089] Based on the established nonlinear state-space model, the state variables are selected as follows: ; in =50- L This represents the amount of wear. The state equation is in discrete form ( =1min): ; The observation equation combines the state variables with the observation vector. H Related.

[0090] Model parameters were obtained through offline identification: using historical data (including monthly wear records, temperature records, and fault records) of 20 carbon brushes replaced over the past 3 years, maximum likelihood estimation was employed. =1.2×10 -5 mm / (A·m·min); =0.3mm / min, =0.15; =0.05mm / min; =0.02 / (℃ min); =1.5×10 -6 / (W·min); =0.01 / min; =0.03V / (mm·N·min); =0.1V / min; =0.02 / min.

[0091] Unscented Kalman filtering (UKF) is used for state estimation. The algorithm flow is as follows: Figure 6 As shown. k =0 state at time 0 Covariance Initialization. Every minute, when a new observation vector... Upon arrival, UKF performs the following actions: Calculate 2 n+1 = 7 Sigma points. Predicting Sigma points using the state equation yields the predicted values. Covariance The predicted Sigma points are mapped to the observation space using the observation equation, and the Kalman gain is calculated. and utilize actual observations Correction status; To obtain the optimal estimate Covariance .by Taking the time when 100 minutes have passed (i.e., the current moment) as an example, the state estimation result is: This is consistent with observations.

[0092] like Figure 7 As shown, based on the current state estimation The future state trajectory is extrapolated iteratively forward using the state equation (ignoring process noise). The failure threshold is defined as follows: Wear failure ≥40mm (i.e., remaining length ≤10mm); Electrical failure: ≥2.5V; Oxidation failure ≥10.

[0093] The prediction results show that It will reach 40mm in about 12 days, and and The carbon brush will not exceed the limit over a longer period of time, so the remaining lifespan is 12 days. Considering the uncertainty of the model and noise, 1000 predicted trajectories were generated through Monte Carlo simulation (with process noise added), and the 95% confidence interval of the remaining lifespan was obtained as [10, 15] days.

[0094] The result is displayed in the visualization interface. Figure 7 The system displays real-time data: the radar chart on the left shows the current scores of carbon brush #23 across various dimensions (wear 0.4, oxidation 0.6, contact pressure drop 0.3, etc., normalized to 0-1, with 1 being optimal); the Gantt chart on the right shows the remaining life bars for all carbon brushes, with carbon brush #23 marked in yellow (warning status). Simultaneously, the system issued a warning when the remaining life was less than 30 days, alerting operators. The system successfully provided early warning of potential faults, ensuring the safe operation of the unit.

[0095] In summary, this invention provides a method and system for assessing the condition and predicting the lifespan of generator slip ring carbon brushes. The method utilizes acoustic emission technology to capture microscopic friction anomalies, machine vision for automatic edge chipping rating, and temperature field inversion to obtain contact pressure drop. Based on these, a three-dimensional coupled attenuation model is established, combined with a nonlinear filtering algorithm, to dynamically predict the remaining lifespan of the carbon brush. This invention integrates multi-dimensional sensor information and deeply analyzes the degradation mechanism of carbon brushes, thereby enabling high-precision and accurate assessment and prediction of the remaining lifespan of generator slip ring carbon brushes, providing technical support for the operation and maintenance of generator slip ring carbon brushes.

[0096] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Anyone skilled in the art can readily implement the present invention according to the description and above. Any modifications, alterations, or equivalent variations made using the technical content disclosed above are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A method for assessing the condition and predicting the lifespan of a generator slip ring carbon brush, characterized in that, Includes the following steps: S1. Simultaneously collect temperature distribution data, friction acoustic emission signal data, macro vision image data, and auxiliary detection data such as excitation current, remaining carbon brush length, and ambient temperature and humidity during the operation of carbon brush and slip ring. S2. Based on macro visual image data processing and analysis, carbon brush edge features are analyzed to obtain carbon brush edge cracking rating results; S3. Extract time-frequency domain feature parameters based on triboacoustic emission signal data to construct a friction anomaly index characterizing the micro-friction state; S4. Based on temperature distribution data and combined with the electrical and geometric parameters of the carbon brush, the contact pressure drop at the contact interface between the carbon brush and the slip ring is obtained by inversion. S5. By integrating carbon brush edge cracking rating, friction anomaly index, contact pressure drop, carbon brush remaining length, and carbon brush axial temperature gradient and oxidation degree index calculated based on temperature distribution data, a multi-dimensional evaluation index system for carbon brush health status is constructed. S6. Based on a multi-dimensional evaluation index system, establish a three-dimensional coupled decay nonlinear state-space model containing carbon brush wear, oxidation degree, and contact pressure drop; S7. Based on the three-dimensional coupled attenuation nonlinear state-space model, the unscented Kalman filter algorithm is used, combined with a multi-dimensional evaluation index system to recursively estimate the carbon brush health status, and extrapolate to obtain the remaining lifespan of the carbon brush and its confidence interval.

2. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In S2, the processing and analysis of macro visual image data includes image preprocessing, semantic segmentation of the carbon brush region, and quantization of edge morphological features. The quantized features are then input into a classification model to obtain the carbon brush edge cracking rating result.

3. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In S3, the processing of the friction acoustic emission signal data first involves bandpass filtering, then frame division using a sliding time window. The extracted time-frequency domain feature parameters include root mean square value, ring count, peak factor, frequency centroid, and kurtosis. The friction anomaly index is obtained by dimensionality reduction through principal component analysis to obtain the first principal component or by calculating Mahalanobis distance.

4. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In S4, the inversion of the contact pressure drop is to solve the inverse problem of heat conduction, based on the carbon brush axial heat conduction equation: ； in, This is the distance from the contact surface between the carbon brush and the slip ring. This refers to the axial temperature distribution of the carbon brush. The perimeter of the carbon brush cross-section is... This represents the cross-sectional area of ​​the carbon brush. The heat flux density of the heat source inside the carbon brush. The convective heat transfer coefficient is... The thermal conductivity of the carbon brush material. For ambient temperature, The density of the carbon brush material. This refers to the specific heat capacity of the carbon brush material. For time.

5. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In S5, the constructed multidimensional evaluation index system is in the form of a multidimensional vector of carbon brush health status, as follows: ； in, A multidimensional vector representing the health status of carbon brushes. This represents the remaining length of the carbon brush. This is an index of the degree of carbon brush oxidation. This refers to the contact pressure drop at the interface between the carbon brush and the slip ring. The axial temperature gradient of the carbon brush over continuous time. This is the friction anomaly index. This refers to the carbon brush edge chipping level. For time.

6. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In S6, the discrete-time state equation of the three-dimensional coupled decay nonlinear state-space model is: ； in, This refers to the amount of carbon brush wear. The degree of carbon brush oxidation, This refers to the contact voltage drop between the carbon brush and the slip ring. For excitation current, The linear velocity of the slip ring. This refers to the carbon brush edge chipping level. This refers to the core temperature of the carbon brush. As the reference temperature, This represents the total length of the carbon brush. For spring pressure, The initial pressure of the spring. The sampling interval is... For discrete-time indexing, The parameters to be identified in the model This is process noise.

7. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In S6, the observation equation of the three-dimensional coupled decaying nonlinear state-space model is: ； in, For discrete-time indexing, This represents the remaining length of the carbon brush in discrete time. This is an index of the degree of carbon brush oxidation over discrete time. These are the observed contact pressure drops between the carbon brush and the slip ring over discrete time periods. The axial temperature gradient of the carbon brush in discrete time. This represents the friction anomaly index in discrete time. The carbon brush edge chipping level is defined as a discrete-time condition. This represents the carbon brush wear over discrete time intervals. The degree of carbon brush oxidation at discrete time. The contact voltage drop between the carbon brush and the slip ring is given in discrete time. The excitation current is in discrete time. It is a nonlinear function. To observe noise.

8. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In step S7, when recursively estimating the carbon brush health state using the unscented Kalman filter algorithm, Sigma points are first generated based on the current state estimate and covariance. The calculation formula is as follows: ； in, for The initial Sigma point at time t, for The first moment Sigma points, for Estimated carbon brush health status at any given time. Let be the state dimension. For scale parameters, for The covariance matrix at time t, The first square root of the matrix List.

9. The method for evaluating the condition and predicting the lifespan of generator slip ring carbon brushes according to claim 1, characterized in that, In S7, the extrapolation of the remaining lifespan of the carbon brush includes: comparing the predicted state trajectory with a preset failure threshold, which includes wear failure, electrical failure, and thermal aging failure. Wear failure is when the carbon brush wear reaches its maximum value, electrical failure is when the contact voltage drop reaches its maximum value, and thermal aging failure is when the oxidation degree reaches its maximum value. The remaining lifespan is obtained by multiplying the difference between the discrete time when the state variable first crosses any failure threshold and the current discrete time by the sampling interval. Multiple predicted trajectories are generated through Monte Carlo simulation to obtain the confidence interval of the remaining lifespan.

10. A system for assessing the condition and predicting the lifespan of a generator slip ring carbon brush, used to implement the method for assessing the condition and predicting the lifespan of a generator slip ring carbon brush as described in any one of claims 1-9, characterized in that, It includes a multi-source heterogeneous data acquisition module, an automatic carbon brush edge chipping rating module, a friction noise feature extraction module, a contact pressure drop and temperature field inversion module, a multi-dimensional evaluation system construction module, a three-dimensional attenuation model modeling module, and a remaining life dynamic prediction module, among which: The multi-source heterogeneous data acquisition module is used to: synchronously acquire temperature distribution data, friction acoustic emission signal data, macro vision image data, and auxiliary detection data such as excitation current, remaining carbon brush length, and ambient temperature and humidity during the operation of carbon brush and slip ring. The automatic carbon brush edge chipping rating module is used to: analyze carbon brush edge features based on macro visual image data processing and obtain carbon brush edge chipping rating results; The friction noise feature extraction module is used to: extract time-frequency domain feature parameters based on friction sound emission signal data, and construct a friction anomaly index characterizing the micro-friction state; The contact pressure drop and temperature field inversion module is used to: invert the contact pressure drop at the contact interface between the carbon brush and the slip ring based on temperature distribution data combined with the electrical and geometric parameters of the carbon brush; The multidimensional evaluation system construction module is used to: integrate carbon brush edge cracking rating, friction anomaly index, contact pressure drop, carbon brush remaining length, and carbon brush axial temperature gradient and oxidation degree index calculated based on temperature distribution data to construct a multidimensional evaluation index system for carbon brush health status. The three-dimensional attenuation modeling module is used to: establish a three-dimensional coupled attenuation nonlinear state-space model containing carbon brush wear, oxidation degree, and contact pressure drop based on a multi-dimensional evaluation index system; The remaining life dynamic prediction module is used to: recursively estimate the carbon brush health status based on a three-dimensional coupled decay nonlinear state-space model, using an unscented Kalman filter algorithm, combined with a multi-dimensional evaluation index system, and extrapolate to obtain the remaining life of the carbon brush and its confidence interval.

Images